Pitch Detection Benchmark

A comprehensive benchmark suite for evaluating and comparing different pitch detection algorithms across multiple datasets and metrics.

📊 Key Findings

Praat emerges as the best choice for pitch detection, offering high pitch accuracy with the lowest cents error across all datasets while also providing unmatched computational efficiency (449x faster than baseline). While pYIN shows higher harmonic mean scores in some datasets, Praat's significantly lower pitch error and dramatic speed advantage make it the clear winner for both accuracy and practical applications. For a detailed breakdown of results, see Benchmark Results.

Pitch accuracy (↑ higher is better)

Algorithm	NSynth	PTDB	MDB-STEM-Synth	Average
YAAPT	56.8%	79.3%	82.8%	73.0%
Praat	72.8%	78.5%	91.7%	81.0%
SWIPE	65.3%	79.6%	90.3%	78.4%
RAPT	67.7%	67.5%	91.7%	75.6%
pYIN	77.8%	81.7%	87.6%	82.4%
TorchCREPE	N/A*	78.2%	N/A*	N/A*
PENN	64.7%	N/A*	N/A*	N/A*

*N/A indicates the algorithm was excluded from testing on this dataset as it was used in training

Cents Error (↓ lower is better)

Algorithm	NSynth	PTDB	MDB-STEM-Synth	Average
YAAPT	1621.25	103.48	392.33	705.69
Praat	499.38	86.76	105.75	230.63
SWIPE	953.01	99.67	181.01	411.23
RAPT	1166.45	109.97	152.39	476.27
pYIN	685.57	131.40	168.22	328.40
TorchCREPE	N/A*	89.17	N/A*	N/A*
PENN	1221.15	N/A*	N/A*	N/A*

*N/A indicates the algorithm was excluded from testing on this dataset as it was used in training

🚀 Quick Start

Installation

pip install torchcrepe crepe praat-parselmouth pysptk librosa penn AMFM-decompy

Basic Usage

Visualize algorithm comparisons:

python visualize_algorithms.py audio_file.wav

Run speed benchmark:

python speed_benchmark.py

Run pitch detection benchmark:

python pitch_benchmark.py --dataset DATASET_NAME --data-dir DATA_PATH

🛠️ Features

• Comprehensive evaluation across standard datasets:
  • PTDB
  • NSynth
  • MDB-stem-synth
• Performance benchmarking for CPU and GPU execution
• Visualization tools for algorithm comparison
• Implementation of popular pitch detection algorithms:
  • YAAPT (pYAAPT implementation)
  • Praat (Parselmouth implementation)
  • TorchCREPE (PyTorch implementation of CREPE) and CREPE (original implementation)
  • Pitch-Estimating Neural Networks (PENN)
  • SWIPE (SPTK implementation)
  • RAPT (SPTK implementation)
  • pYIN (librosa implementation)

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

📚 Citation

If you use this benchmark in your research, please cite:

@software{pitch_detection_benchmark,
  title = {Pitch Detection Benchmark},
  author = {Lars Nieradzik},
  year = {2025},
  url = {https://github.com/lars76/pitch-detection-benchmark}
}